Sensor Webs
An Emerging Concept for Future Earth Observing
Systems
An EOS Brown-Bag Lunch Presentation
Stephen J. Talabac
NASA/GSFC Code 586
April 11, 2003
Agenda
Background, terminology, and fundamental concepts
A candidate sensor web definition
Taxonomy and properties of sensor web nodes
Possible sensor web classes and their properties
Representative scenarios that may benefit from the
sensor web concept
A survey of related research activities
11 April 2003
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 2
Background
“The best way to be ready for the future is to invent it.”
John Sculley – CEO, Apple Computer
NASA’s and the Earth-Science Enterprise’s strategic plans
identify “sensor webs” as a new paradigm for conducting future
science observations.



“We envision multiple cooperative spacecraft that operate in
interactive networks to thoroughly explore diverse phenomena”
“…intelligence will become an integral part of future spacecraft,
enabling systems to make real-time decisions in the uncertain and
unforgiving space environment”.
“Deploy cooperative satellite constellations and intelligent sensor
webs.” that will facilitate“information synthesis” and increase our
“access to knowledge”
11 April 2003
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 3
Background (continued)
The Sensor Web concept is being refined and various views of it
appear to be converging.
We are presently exploring questions such as:




What exactly “is” a sensor web?
What are the specific characteristics that a sensor web should
possess?
What are the various behaviors that a sensor web may manifest?
…and most significantly: What are potential applications of sensor
webs that can be of significant benefit to the Earth science
community?
11 April 2003
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 4
Sensor Webs:
A Systems Engineering Approach
Establish a common terminology vocabulary
Identify sensor web nodes, define their properties,
and develop a node taxonomy
Describe how nodes might interact and used as
building blocks to develop a taxonomy of sensor web
classes
Identify science scenarios that may benefit from the
various sensor web classes
Identify candidate sensor web architectures and
establish evaluation criteria
11 April 2003
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 5
Space Mission Architecture - Today
Direct instrument
readout
Bent pipe
communications
or
On-board recorder
downlink to Ground
Station
Science Processing Center
Science Processing Center
Graphic Credit: NASA/GSFC 2000 Survey of Distributed Spacecraft Technologies and Architectures
for NASA’s Earth Science Enterprise in the 2010-2025 Timeframe
11 April 2003
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 6
Space Mission Architecture - Today
Classic “stovepipe1” science data collection and mission operations
Single or separate spacecraft missions with little or no dynamic
planning for opportunistic science observations or handling unexpected
observing conditions
Data is often simply recorded and downlinked to ground systems for
processing and analysis


“Fire hose” of raw data bits downlinked to the ground with little or no
regard to sending just the most meaningful science data
Little, if any, on-board science instrument processing
No real time collaborative information sharing between sensors,
spacecraft, or investigators
Interspacecraft communications typically relegated to bent pipe
communications to the ground segment

1
e.g., via TDRSS in support of command uplinks, telemetry downlinks
stovepipe - a self-standing, narrowly focused application that solves a discrete set
of problems
11 April 2003
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 7
Space Mission Architecture – “Tomorrow”
Distributed Spacecraft Systems & Sensor Webs
Graphic Credit: NASA/GSFC: 2000 Survey of Distributed Spacecraft Technologies
and Architectures for NASA’s Earth Science Enterprise in the 2010-2025 Timeframe
11 April 2003
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 8
Space Mission Architecture - “Tomorrow”
High degree of synergy between a diverse suite of platforms




Space-based
Atmospheric (e.g., aircraft, balloons)
Land (e.g., in-situ weather stations) and sea (e.g., buoys)
Subsurface probes
Automated science data collection and mission operations
Multiple spacecraft and platforms perform dynamic planning for
opportunistic science observations
Real time collaborative information sharing between sensors,
spacecraft, or investigators
Interspacecraft communications becomes an intrinsic
characteristic of distributed space platforms
11 April 2003
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 9
Emerging & Evolving Technologies
Micro-electromechanical Systems
(MEMS)


MEMS Microthruster Array - 10-4 Ns Impulse
Photo : TRW
Nanospacecraft
Low mass, small footprint in-situ
sensors
Advanced processors & high
capacity storage provide


Greater opportunity for on-board
processing
Embedded software to build
“intelligent” processing nodes
Communications



Evolving space-based IP comms
protocols and interoperability with
terrestrial networks
Ubiquitous wireless comms allows
for “ad hoc” Sensor Web networks
to be established
Peer-to-peer networking
11 April 2003
NASA NMP - ST5 spacecraft
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 10
Terminology1 - Sensor
A device that measures a physical property of a natural or man-made
phenomena that is of interest to Earth- and space-scientists


A sensor is an integral part of, and provides its measurements to, a science
instrument.
Examples
 A prism and CCD sensor array that captures and measures photons of different
energies (i.e., wavelengths) and intensities (i.e., photons per second)
 A water flow rate sensor

Two types of sensor measurements
 in-situ (e.g., magnetic field strength)
 remotely sensed (infrared energy reflected from clouds at different heights)

Measurements are not necessarily restricted to photons




Rainfall amount
Acoustic energy
Chemistry
1D/2D/3D/6DOF directional measurements
Note 1: These are not meant to be strict definitions. Instead they are intended to provide an
“informal” definition of fundamental concepts, often expressed in terms of other widely
accepted and commonly understood terminology.
11 April 2003
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 11
Terminology - Science Instrument
A self-contained infrastructure that




Receives (digital or analog) data from one or more sensors
Provides temporary storage for the measurements
Transmits sensor data to a node
It may have its own infrastructure to sustain its own operation
(e.g., power, structure, environmental, etc), or it may rely entirely
on the node (e.g., a spacecraft bus) for this infrastructure
Examples




rain gauge
magnetometer
IR observatory
laser interferometer
11 April 2003
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 12
Terminology - Node
A self-contained computing, storage, and communications device.

A node may, but not necessarily, be connected to one or more science
instruments.
A node provides mechanical, power, thermal, electrical,
communications, control, timing, environmental protection, etc. to
support its own operation and possibly for any science instrument
directly connected to the node.


A node may be a spacecraft that has one or more science instruments
A node may be a computing or storage system that does not have any
associated science instrument(s)
Example of nodes:




A spacecraft bus that supports one or more science instruments
A ground-based radio telescope observatory
A computer that executes a model (e.g., numerical weather prediction; algal
bloom formation, growth, and dispersion)
A geolocated database that stores historical information (e.g., seasonal
hurricane formation locations; seasonal algal bloom population emergence
locations)
11 April 2003
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 13
Node Types
Instrument Nodes
Computing/Data Storage
Nodes
Integrated Instrument
Stand-alone Instrument
Sensors
Sensors
Science
Instrument(s)
Science
Instrument(s)
No Science Instrument
Node
Communications
Fabric interface
Science Instruments are “tightly
coupled” to a node (e.g., a
spacecraft bus)
11 April 2003
Node
Node
Communications
Fabric interface
Communications
Fabric interface
Science Instruments are “loosely
coupled” to a node (e.g., a ground
observatory linked via
communications link to a remote
ground data system)
A computing node is not
connected to any science
instrument (e.g., a meteorological
forecast model running on a
computer system)
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 14
Node Concept & Connectivity
Another view…
Science Instrument
&
Sensors
Optional Instrument with Sensors
Mechanical
Node Platform
Sensor m Sensor 1 Sensor 2
Sensor 1 Sensor 2
Power
Thermal
Node
Control
Sensor n
Optional Instrument with Sensors
Instrument Instrument
Data
Data
Processor
Storage
Comms
Fabric
I/F
Node Platform
Communications
Fabric
This node does not have
a Science Instrument/Sensors
11 April 2003
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 15
Terminology - Data
Loosely defined to mean any “string of bits”.
Data bits may represent







Raw sensor or science instrument data
Processed science data
Ancillary information required to perform science data processing
Node or instrument state data (e.g., spacecraft health and safety
engineering telemetry data; instrument mode of operation; …)
Commands to the node and/or science instrument to change its
operating state
Executable code (i.e., algorithms) to be executed by the node
Sensor web system state messages
11 April 2003
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 16
Terminology -
Communications Fabric
A communications infrastructure that permits nodes
to transmit and receive data between one another
The scope of the communications fabric
encompasses



The communications media (e.g., wired vs. wireless; optical
vs. RF; baseband signal vs. modulated; etc)
The communication topology (e.g., ring, star, mesh, etc)
Communications fabric protocols
11 April 2003
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 17
So...what “is” a sensor web?
“Scientific progress consists in the development of new concepts.”
Ernst Mayr – renowned 20th century evolutionary biologist
As is often the case with emerging concepts, there is presently no
single, widely accepted definition.
A candidate definition:
A Sensor Web is a distributed system of sensing nodes that are interconnected by
a communications fabric and that functions as a single, highly coordinated,
virtual instrument. It autonomously detects and dynamically reacts to events,
measurements, and other information from constituent sensing nodes and from
external nodes (e.g., predictive models) by modifying its observing state so as
to optimize science information return.”
11 April 2003
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 18
Sensor Web Conceptual Diagram
Database Node
Information Fusion/Synthesis Computing Node
Instrument Node
Instrument Node
Science instrument
photons
Science instrument
photons
Communications
Fabric
Predictive Model Node
(has no science instrument)
Non-photon science instrument
data
(e.g., distance measurements
from a laser rangefinder)
Instrument Node
Notes:
(1)
The communications fabric is not meant to imply just one “network”, nor does it imply any particular
medium (RF, wired, fiber optics), nor specific connectivity such as a ring network versus a fullyconnected mesh topology. It simply means that nodes use the communications fabric to send and
receive data to/from one another.
(2)
Some nodes are shown having science instruments whereas others do not have instruments. An
example of a node with no instruments is a computer system that executes a numerical
meteorological forecast model and that provides its results to one or more other nodes.
11 April 2003
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 19
Sensor Web Components:
Autonomous Dynamic Interactions
• In situ & remote sensing observations
• Individual & collaborative event
detection and phenomenon recognition
Sensor
Nodes
• Notification of other nodes
• Reaction & Response by nodes
• Node reconfiguration
Communications
Fabric
Computing
Nodes
Data
Stores
• Temporal (e.g., measurement rate),
spatial (e.g., new location, higher
resolution, form new cluster),
spectral (e.g., activate different
band)
e.g., Predictive models, e.g., Historical Information
information synthesis,
observational data assimilation
11 April 2003
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 20
Why are Sensor Webs Important?
“In the future, the research models of today will be the application models
of tomorrow… What kind of observing system will we need?”
Dr. Mark Schoeberl, GSFC Science Data Processing Workshop February 2002
“Sensor-web enabled systems are
uniquely capable of performing realtime analysis and decision making
to autonomously execute complex
adaptive observing strategies.” E.
Sensor webs will inextricably link insitu & remotely sensed observations
with model outputs and information
repositories from geographically
dispersed and disparate sources; not
possible with stand-alone sensors.
“Improve the performance of weather
and climate predictive systems and
extend useful range of forecasts.” Ibid
Torres-Martinez, M. Schoeberl, M. Kalb;
June 2002 IGARSS
Ability to “aggregate/synthesize
science data by clustering, or some
other local data aggregation
methods, to generate global highlevel interpretations” XEROX PARC
CoSense Project
Advanced
Sensors
E. Torres-Martinez et al
Sensor
Webs
Information
Synthesis
Access to
Knowledge
Graphic Credit:
Earth Science Vision 2002
Access to Knowledge
Dr. M. Schoeberl
11 April 2003
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 21
Why are Sensor Webs Important?
“Ideas won't keep; something must be done about them.”
Alfred North Whitehead
Achieve science objectives
unattainable using single nodes

Phenomena that occupy a very
large spatial domain - “Local
data is too weak to form
coherent global interpretation”
Reduce system response time


 Xerox PARC “CoSense” project
 Magnetosphere


Multipoint, time synchronous
observations
Arrays of large effective
aperture instruments


Monitor rapidly evolving,
transient, or variable
events/phenomena
Conduct time constrained
observations without a priori or
having incomplete knowledge of
conditions at observing time
Conduct observations where
communication times are too
long for humans to make realor near-real-time decisions
Improve utilization of platform &
instrument resources
The phenomena we observe are intrinsically dynamic … as must be the
sensor web information systems that will enhance our ability to observe and
better understand these phenomena.
11 April 2003
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 22
A Groundswell of Sensor Web Research
Multi-Resolution Data Fusion
Duke University - SensIT
Collaborative Sensemaking of
Distributed Sensor Data
DARPA: Dynamic Sensor Networks
Credit: XEROX PARC, DARPA
Credit: NASA/JPL
Credit: DARPA “SensIT“ project
Simulating highly scalable routing
protocols for 10,000 node sensor
networks.
Sensor Web for In Situ Exploration
of Gaseous Biosignatures
UAV Research
Berkeley SensIT
11 April 2003
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 23
Sensor Web Research
NASA/JPL
“A Sensor Web consists of intracommunicating, spatiallydistributed sensor pods that are
deployed to monitor and explore
environments.”
“It is capable of automated
reasoning for it can perform
intelligent autonomous operations
in uncertain environments,
respond to changing
environmental conditions, and
carry out automated diagnosis
and recovery.” Dr. Kevin Delin, JPL
A Sensor Web measuring biosignature gases to
search for microorganisms living beneath the
surface of a planet.
Images Credit:
NASA/JPL
Sensor Web project leader
The "hopped" data is shared by
all of the pods, allowing each one
to know what is being collected
elsewhere on the web.”
11 April 2003
JPL Sensor
Web “Pod”
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 24
Sensor Web Research
DARPA Sensor Information Technology Program
Xerox/PARC: CoSense Project



Using sensor collaboration to make
sense of aggregate phenomena that
are not local in time and space
Leader-follower formations
Clustering of enemy forces
Track multiple maneuvering targets,
without a priori knowledge of paths


A data association problem
Estimate target position versus time
How many nodes are required?


What are the impacts of node
spacing?
How to perform distributed analysis
with 100s and 1000s of nodes
For very large numbers of nodes

How to perform (and perhaps
optimize) cluster maintenance and
node reconfiguration to ensure
efficient node collaboration?
11 April 2003
Perimeter violation sensing

monitor only events within a
predefined area; ignore all others.
Target tracking and “reasoning”

Detecting a particular target
signature implies the existence of
another target signature
Information directed sensor
querying

Select next sensor to query to
maximize information return while
minimizing latency & bandwidth
consumption
How do you infer the properties of a
global set of targets vs. individual
target properties?
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 25
Sensor Web Research
DARPA Sensor Information Technology Program
Sensoria Corporation

Examining dynamic network
assembly to build
deterministic networks
BAE and Sensoria



Auto track a vehicle
Create initial estimate of
future velocity and location
Coordinate all nodes to
image the vehicle when it is
in view
Nodes share event detection
info and tracking states
Nodes contribute to improved
initial tracking estimate
Technologies


 Seismic
 Acoustic
 Infrared


11 April 2003
Sensor signal processing
Sensor signal/information
fusion
Routing algorithm is insensitive
to loss of nodes
Image trigger estimators
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 26
Sensor Web Research
DARPA Sensor Information Technology Program
MU-Fashion


Multi- Resolution Data Fusion using Agent- Bearing Sensors In
Hierarchically-Organized Networks
Duke Univ, LSU, Univ. of TN
Examining problems associated with:




Sensor data fusion
Multi-resolution, fault tolerant target detection & classification
Sensor deployment algorithms to optimize target detection and minimize
communications bandwidth
Working with BBN on sensor power management
 for devices that operate in multiple power states
Using RT-Linux
11 April 2003
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 27
Sensor Web Research
DARPA Sensor Information Technology Program
BBN


Large-scale, distributed, intelligent sensor networks
10,000 nodes +
 Use peer-to-peer communications protocols
 Ad hoc mobile networks

Scalable architecture
 supports numerous diverse and heterogeneous sensor types

XML Messaging Standards
 allows sensors to communicate and share information


XML messaging standards
Java-based solution
11 April 2003
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 28
Sensor Web Research
DARPA Sensor Information Technology Program
Dynamic Sensor Networks USC, UCLA, VA Tech
Distributed Services for Self
Organizing Sensor Networks -
Distributed Cognition through
Semantic Information Fusion:
Penn State

Auburn University

“Provide services that enable
distributed sensor software
components to self-organize,
adapt to changing
requirements, react to network
changes, relocate and survive
sensor failures in a dynamic ad
hoc network”
11 April 2003


Low bandwidth comms requires
abstraction of data
Platforms self organize into
local neighborhoods and share
local data
Collaborative signal processing
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 29
NASA/GSFC
Autonomous Nano Technology Swarm: ANTS
An Artificial Intelligence Approach to Asteroid Belt Resource
Exploration: Dr. Steve Curtis
Scientifically categorize all asteroids > 1 km in diameter

“Mission goals are achieved through emergent, collective behavior”. Dr. Steve Curtis
Very large numbers (“swarms”) of picospacecraft (~1kg) with wide variety of
instruments




X-ray
Gamma ray
Magnetometer
IR/Vis/UV spectrometers
Swarm heuristics planner and distributed intelligence operations
11 April 2003
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 30
Sensor Web Research at GSFC
Sensor Web Application Prototype (SWAP)
ESTO FY01 Funded Prototype
Research



weather
simulator
Simulated
Doppler Radar

Meteorological application in
collaboration with Dr. Marshall
Shepard
Dynamic instrument collaboration
for flash flood prediction
Automated response and
reconfiguration of simulated NWS
Doppler radar array
“Intelligent“ rain gauges
automatically initiate a simulated
Doppler radar mode change from
“sweep“ to “sector scan“.
Doppler Radar “Sector Scan” Command
Embedded Processor
11 April 2003
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 31
Sensor Web Research at GSFC
An Autonomous, collaborative in situ marine fleet observing
system (OASIS)
SolarArrays
Arrays
Solar
Dynamically control fleet sampling
strategy to observe a cold core eddy as it
develops from the Gulf Stream and then
sheds into the subtropical gyre.
Potential use in real-time data
assimilation efforts
Harmful Algal Blooms
Jet Discharge
Sensor Suite:
- Microsalinograph; Fluorometer; Radiometers
- Wind Monitor; RH/Temperature Probe
- Precision Barometer; GPS
Monitor their growth, map boundary,
conduct in situ measurements
Batteries
Propulsion Tubes

Jet Intake
Payload Tube
Power: Marine solar panels & marine batteries
Iridium: 2-way real-time communications
Control: Twin thrusters with internal rudder
OBC: G&N and sensor control
Planned capabilities: Grid mapping, dynamic
surveying, and station-keeping.
Dynamic Ocean Processes: Meanders,
Frontal Instabilities,eddies
11 April 2003
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 32
OASIS Platform Sensor Web
Gulf Stream Eddy Mapping and Nutrient Measurement
Eddy Properties (e.g., boundary)
Coordinate in situ and space based observations
“Cold rings trap nutrient-rich
water and transports nutrients
and plankton into the
relatively-barren Sargasso Sea”
Credit: Gulf of Maine Aquarium
Credit:
home page
11 April 2003
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 33
A large algal bloom…
…could a GSFC OASIS sensor web fleet have helped?
Commercial fishermen along the Southwest Florida
coast are reporting a massive dead zone that is almost
devoid of marine life in an area of the Gulf of Mexico
traditionally known as a rich fishing ground.
They've dubbed it black water, and they're demanding
that local, state and national government agencies find
out what's causing it.
Scientists who have heard of the phenomenon say
they, too, need answers.
"It's killed a lot of the bottom because recently a lot of
little bottom plants are coming to the surface dead and
rotten out in the Gulf," said Tim Daniels, 58, a
Marathon Key fish-spotting pilot who has been flying
over the Gulf for more than 20 years.
Like Daniels, fishermen with decades on the water say
they've often seen red tide but they've never seen
anything like this — it doesn't have a foul smell, it isn't
red tide and it isn't oil. They describe it as viscous and
slimy water with what looks like spider webs in it.
Credit: NASA Earth Observatory - SeaWiFS on Orbview-2
11 April 2003
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
March 17, 2002
Slide 34
Another Potential Sensor Web Application
Human Health and the Environment
Fixed and mobile sensors are
placed in areas where humans
cannot go or because it is too
dangerous

e.g., Exxon-Valdez oil spill
Areas particularly vulnerable to
oil spills


Mud flats: can be up to several
kilometers wide in the Cook
Inlet/Kenai Peninsula region
Can be dangerous environment
for humans to work in
A potential sensor web solution?


A fleet of autonomous
amphibious vehicles
Perform collaborative mapping
and contaminant measurements
on exposed tidal/mud flats
11 April 2003
Sediment plume resulting from oil cleanup
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 35
GSFC Sensor Web Concept Formulation
Developed candidate
sensor web definition
Characterizing





Science needs
Candidate science
scenarios and
applications
Developing


Prototypes
Models and simulations
11 April 2003
GIS Terrain Database
Science Data Archive
Storage
database
Platform
Processors
Computing/Data Storage
Platforms
Platform
ƒ(x)
Platform
Collector-Reactor
Reactor-Processor
Collector-Processor
Passive Collector
Collector-Reactors
Active Collector
Instrument
Identifying
ƒ(x)
Science Data
Orbit Determination
Processor
Processor
Plan & Sched.
Processor
Instrument

Node taxonomy and
properties
Node data reporting
properties
Sensor web classes
Required sensor web
architectural properties
Platform Types & Taxonomy
Collectors
M(i)
ƒ(x)
Instrument Platforms
Platform
Sensor Web Properties
Node Aggregation
Cluster 1 Cluster 2
Cluster 1
Cluster 1 Cluster 2
Cluster 2
Clustering, Reconfiguration, & Reassignment
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 36
Node Properties
Location & Reporting Modes
In space

A single or a Distributed Space
System (DSS) mission
Within Earth’s (or planet’s)
atmosphere

Aircraft (e.g., UAV), balloon,
sounding rocket...
On the Earth’s (or planet’s)
surface


Fixed and mobile nodes
Weather station, autonomous
land/water craft
Beneath the planetary surface
or “skin”

Submarine, subsurface sensors
11 April 2003
Deterministic


Node outputs information at a
priori known times
This does not necessarily mean
a “fixed” time interval
Triggered

Node reports information only
when a predefined event,
condition, phenomenon, or
specified node operating state
is detected
On-Demand

Node reports information when
requested by another node
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 37
Class 1 Nodes
Collectors
A science data collector


Has n well defined distinct states (modes) of performing data
measurements (i.e., n data collection modes)
Mode mi of mn available modes may be selected but not modified
Transmits raw instrument data only

Is unable to receive, and therefore cannot react to, data that is
transmitted by another node
Three Class 1 node types



Passive Collector
Active Collector
Collector-Processor
11 April 2003
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 38
Class 1 Node
The Passive Collector
Input: collects raw science
instrument measurements
Data Processing: None



Node examples:


One data collection mode
Simply formats raw data for
communications fabric
transmission compatibility
No science instrument state
changes supported
Tipping bucket rain gauge
Panchromatic, multispectral
downlinked raw data stream
with only one resolution, fixed
FOV, etc.
Output: raw science instrument
measurements and
instrument/node state data
11 April 2003
Instrument
Raw Sensor Data
Raw Instrument
and Node State Data
Node
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 39
Class 1 Node
The Active Collector
Input: collects raw science
instrument measurements
Data Processing: Some


Node examples:

Data collection mode changes
are supported
Science instrument mode (M)
selection by the node using a
simple function: M(i)
dw/dt
 Serves to maximize available
comms bandwidth efficiency
 Send information that is most
suitable to the sensed condition
Output: raw science instrument
measurements and
instrument/node state data
11 April 2003
Instrument
Raw Sensor Data
River gauge node reports water
level (w) measurements at more
frequent time (t) intervals when
node’s M(i) detects increasing

Spacecraft selects a sensor best
suited to phenomena of interest
(e.g., NIR vs. visible vs. UV)
Raw Instrument
and Node State Data
M(i)
Node
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 40
Class 1 Node
The Collector-Processor
Input: collects raw science
instrument measurements
Data Processing:


Node examples:

Fixed algorithm supported
Instrument mode changes are
supported as with Active
Collector node
Raw Sensor Data
11 April 2003
Instrument
Raw Sensor Data
ƒ(x)
Instrument
Output: processed science
instrument measurements and
instrument/node state data
DMSP imager: records full
resolution visible pixels and
reduced resolution (averaged)
IR pixels during day time
passes and vice versa during
nighttime passes
M(i)
ƒ(x)
Processed Instrument
and Node State Data
Node
Processed Instrument
and Node State Data
Node
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 41
Class 2 Nodes
Reactors
Collect and react to data sent to it from a science
instrument or from another node

e.g., water level gauge that can report data at one of N
possible time intervals (1x per minute, 4x per minute, etc)
Instrument and/or node has more than one mode of
operation (i.e., data collection and/or processing)
Two class 2 node types


Collector-Reactor
Reactor-Processor
11 April 2003
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 42
Class 2 Node
The Collector-Reactor
Input:


Node examples:
Collects raw science instrument
measurements
Collects data from other
node(s)


Data Processing: None
Output:

Reacts to data/commands from
other nodes and outputs raw
science instrument
measurements
Commands & Data
11 April 2003
Instrument
Raw Sensor Data
River gauge reports water level
(w) measurements at time (t)
intervals
In contrast to the ActiveCollector node, another node
detects a dw/dt condition and
commands a data collection
rate change in the CollectorReactor node
Raw Instrument
and Node State Data
Node
Commands & Data
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 43
Class 2 Node
The Reactor-Processor
Input:


Node examples:
Collects raw science instrument
measurements
Collects data from other nodes

 Numerical forecast model (A
Science Processor class node)
forecasts severe storm and
commands GOES-x spacecraft
(A Reactor-Processor Node) to
begin rapid scan highresolution imaging mode
Data Processing:

Reacts to input data/cmds and
outputs processed data
Output:

Processed science data and
node state data
Commands & Data
11 April 2003
Instrument
Raw Sensor Data
Future GOES-x meteorological
spacecraft
ƒ(x)
Node
Processed Instrument Data
and Node State Data
Commands & Data
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 44
Class 3 Node
Processors & Data Storage
Are not connected to a science instrument
Input data from one or more other nodes
Processor nodes transform received data into one or more
higher level products
Output stored data or processed science data products



meteorological forecast model results
remapped imagery
synthesized science data (e.g., multispectral imagery and SAR
combined to produce 3D terrain map)
There may be many node types in this class




Science processor
Database, science archive...
Orbit and attitude processor
Scheduling and planning processor
11 April 2003
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 45
Class 3 Node
The Science Processor
Input:

Node examples:
Collects data only from other
nodes
Output:



Outputs processed science data
A meteorological numerical
forecast model
A MHD model of the
magnetosphere
 Processed science data
 Model or simulation results
Data Processing: Yes
ƒ(x)
Node
11 April 2003
Processed Data (Science, Model, Simulation)
and Node State Data
Commands & Data
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 46
Class 3 Node
The Data Storage Node
Input:

Node examples:
Collects data only from other
nodes
Output:




Outputs stored data

GIS map database
Archived science data
Historical model runs
Results of Data Mining
Data Processing: Perhaps. (e.g.,
Autonomous Data Mining)
Stored Data
Node
11 April 2003
Commands & Data
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 47
Sensor Web Topologies
Hierarchical
• Centralized command and control
• Perhaps decreasing functionality/capability at
lowest nodes/levels
Fully connected mesh
• Peer-to-peer
• Distributed control
• Equivalent functionality/capability at all nodes
Ring
• Store and Forward
• Conducive to pipelined information processing
= A Sensor Web Node
11 April 2003
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 48
Sensor Web Topologies Master
Cluster 1
11 April 2003
Cluster 2
continued
Clustering
• Local command and control and sensor
web data collection within each cluster
• Reports to a “Master” overall control,
monitoring, and/or coordinator node
• A “cluster” may be a formation flying
mission of N spacecraft; or subgroups of
robotic explorers
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 49
Sensor Web Topologies
Reconfiguration – Node Aggregation
Master
Master
Sensor Web at time t1
New Platforms Are
Added at time t2
11 April 2003
And the sensor web
“aggregates” into a new
larger sensor web
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 50
Sensor Web Topologies
Reconfiguration – Failure and Recovery
Master
Master
Master
Master
Master Node Fails
“Drone” Platforms Temporarily
Assume Master Roles
11 April 2003
Then drones negotiate
control and recombine to
a new topology
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 51
Sensor Web Topologies
Reconfiguration – Cluster Mission Reassignment
Master
Master
Cluster 1
Cluster 1
Cluster 2
Cluster 2
Master
Master
Cluster 1
temporarily “breaks
away” and operates
independently
11 April 2003
Cluster 1
Cluster 2
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Then Cluster 1
rejoins the
sensor web
Slide 52
Sensor Web Topologies
Reconfiguration – Cluster Reconfiguration
Master
Master
Cluster 1
Newly formed Cluster 1
Cluster 2
Master
Cluster 1 nodes
temporarily “break
away” and operate as
independent nodes
11 April 2003
Independent
Platforms
Then regroup to
form a new Cluster
Cluster 2
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 53
We can now build a range of Sensor Web Classes
from these nodes types
Class 1
 Event detection and
notification (e.g., Swift GRB and
the Gamma Ray Burst
Coordinates Network - GCN)
 Collaborative system level
reconfiguration (e.g., run new algorithm
based on node data collected; deploy
additional sensors/interrogate other
sensors to refine forecast and alerts)
Simple
Complex
 Data collection and
reporting (e.g., river water
levels, incident solar radiation,
lightning detection network,
GOES DCS for data
dissemination)
 Data sharing (e.g., GPSbased time and location
dissemination)
11 April 2003
Class 3
 Coordinated data collection
and reporting, science data
fusion, information synthesis,
and decision making.
Class 2
 Intelligent sensor web
predicts phenomenon, controls
resources to perform
collaborative observations,
possibly even takes actions to
modify effects of phenomena
(e.g., predicts flash flood;
monitors rainfall conditions;
accesses GIS data; opens flood
gates when flash flood detected)
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 54
Sensor Webs: Architectural Properties
"Life was simple before World War II. After that, we had systems.”
Rear Admiral Grace Murray Hopper
Heterogeneity


Node types and reporting modes
Communications fabric
implementations
Scalability



Reliability, Recovery, Reconfiguration



Sensor Web nodes will almost
certainly be required to
“aggregate” over time
Node population growth should
not adversely affect sensor web
functional or performance
characteristics
Logically combine two or more
sensor webs into one virtual
sensor web

Topologies



Centralized vs. decentralized
peer-to-peer
Local clusters
11 April 2003
Hierarchical, fully connected mesh,
others…
Clustering
Data Management and Delivery


Sensor Web Control

Handle failures, degradation
Platforms: add, subtract, replace,
upgrade/new functions
Mobile nodes
Accommodate changes in
relationships of nodes

Support data and services “discovery”
Semantic and syntactic “seamless”
data/information exchange
Metadata representation & exchange
Sensor Web Security
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 55
Future Sensor Web Research
Sensor Webs must evolve from a strategic plan concept to well
defined science scenarios that can benefit from this new form of
observing system and a mature suite of new information
technologies with which they can be successfully implemented




GSFC Earth scientists and information technologists are collaborating
to identify and describe candidate science scenarios where sensor
web information technologies will yield significant benefits to the
science community
We plan to develop additional Sensor Web prototypes to evaluate
pragmatic information technology implementation issues
Planning to use simulations and modeling techniques to assess
candidate collaborative observing strategies
Leverage existing technologies, infuse emerging new information
technologies, and make investments to maximize the return of useful
science
11 April 2003
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 56
So..where do we go from here?
GSFC is seizing the opportunity to realize a Sensor Web vision
and develop a viable capability for new Earth science
observations, discovery, and understanding.
Links to web resources:


http://pioneer.gsfc.nasa.gov/public/sensorweb/current_research.htm
Google searches on “Sensor Networks”
Contact:

stephen.j.talabac@nasa.gov
11 April 2003
GSFC Sensor Web Concept Formulation – Stephen J. Talabac
Slide 57